The I Theory of Money

Markus K. Brunnermeier

and Yuliy Sannikov

rst version: Oct. 10, 2010

this version: June 5, 2011
Abstract
This paper provides a theory of money, whose value depends on the functioning of
the intermediary sector, and a unied framework for analyzing the interaction between
price and nancial stability. Households that happen to be productive in this period
nance their capital purchases with credit from intermediaries and from their own
savings. Less productive household save by holding deposits with intermediaries (in-
side money) or outside money. Intermediation involves risk-taking, and intermediaries
ability to lend is compromised when they suer losses. After an adverse productivity
shock, credit and inside money shrink, and the value of (outside) money increases,
causing deation that hurts borrowers even further. An accommodating monetary
policy in downturns can mitigate these destabilizing adverse feedback eects. Lower-
ing short-term interest rates increases the value of long-term bonds, recapitalizes the
intermediaries by redistributes wealth. While this policy helps the economy ex-post,
ex-ante it can lead to excessive risk-taking by the intermediary sector.

Preliminary and Incomplete. We are grateful to comments by seminar participants at Princeton, Bank
of Japan, Philadelphia Fed, Rutgers, Toulouse School of Economics, Wim Duisenberg School, University of
Lausanne, Banque de France-Banca dItalia conference, University of Chicago, New York Fed, Chicago Fed,
Central Bank of Chile, Penn State, Institute of Advanced Studies, Columbia University, University of Michi-
gan, University of Maryland, Northwestern, Cowles General Equilibrium Conference, Renmin University,
Johns Hopkins, Kansas City Fed, IMF, LSE, LBS, Bank of England and the Central Bank of Austria.

Princeton University.

Princeton University.
1
1 Introduction
A theory of money needs a proper place for nancial intermediaries. Financial institutions
are able to create money, for example by lending to businesses and home buyers, and accept-
ing deposits backed by those loans. The amount of money created by nancial intermediaries
depends crucially on the health of the banking system and on the presence of protable in-
vestment opportunities in the economy. This paper proposes a theory of money and provides
a framework for analyzing the interaction between price stability and nancial stability. It
therefore provides a unied way of thinking about monetary and macroprudential policy.
Since intermediation involves taking on some risk, a negative shock to productive agents
also hits intermediary balance sheets. Intermediaries individually optimal response is to
lend less and accept fewer deposits. Hence, the amount of inside money in the economy
shrinks. Because money serves as a store of value and the total demand for money changes
little, the value of outside money increases inducing deationary pressure. More specically,
in our model the economy moves between two polar cases: in one case the the nancial
sector is well capitalized: it can overcome nancial frictions and is able to channel funds
from less productive agents to more productive agents. Financial institutions through their
monitoring role enable productive agents to issue debt and equity claims. The value of money
is low. In contrast in the other polar case, in which the nancial sector is undercapitalized,
funds can only be transferred via outside money. Whenever an agent becomes productive he
buys capital goods from less productive households using his outside money, and vice versa.
That is, outside money allows de facto some implicit borrowing and lending. In this case
outside money steps in for the missing nancial intermediation. The value of money is high.
A negative shock to productive agents lowers the nancial sectors risk bearing capacity and
hence brings us closer to the second case with high value of money. That is, a negative
productivity shock leads deation a la Fisher (1933). Since nancial institutions accept
demand deposits they are hit on both sides of their balance sheet. First, they are exposed
to productivity shocks on the asset side of the balance sheet. Second, their liabilities grow
after a negative shock as the value of money increases. This amplies the initial shock even
further, absent appropriate monetary policy. This leads to non-linear adverse feedback loop
eects and liquidity spirals as studied in Brunnermeier and Sannikov (2010). In addition,
the money multiplier is endogenous and declines with the health of the nancial system.
We then study the eect of monetary policy on the intermediary sector. We focus on
budget-neutral monetary policies that do not aect government spending, and consider both
2
interest-rate policies (implemented by printing money) and open-market operations that
change the maturity structure of government liabilities. We nd that in the absence of long-
term government bonds, interest rate policy on its own has no real eects on the economy.
That is, if the monetary authority pays interest on reserves by printing money, the only eect
is that the interest rate always equals the nominal ination. This result holds because all
debt is short-term in our model, and we do not allow for surprise changes in the interest-rate
policy.
However, with long-term (government) bonds that pay a xed rate of interest, interest-
rate policy can have real eects. Then, if the monetary authority always sets a positive
interest rate, bonds and cash are substitutes in the sense that unproductive households
can use both of these assets to store wealth.
1
If the monetary authority lowers interest
rates in downturns, then the value of long-term bonds rises. Intermediaries can hold these
long-term nominal assets as a hedge against losses due to negative macro shocks. In other
words, interest-rate cuts can help banks in downturns, as long as they hold nominal assets
with longer maturities. We refer to this phenomenon as stealth recapitalization as it
redistributes wealth. Importantly, this is however not a zero sum game. Of course, while
this policy can help banks ex-post, by reducing further losses that they are exposed to, it
can create extra risk-taking incentives ex-ante.
Related Literature. While in almost all papers, a negative productivity shock causes
inationary pressures, in our setting it induces deationary pressure absent a monetary in-
tervention. This is consistent with the empirical output-ination patterns before 1960 under
the (extended) Gold Standard, as e.g. documented by Cagan (1979). Like in monetarism
(see e.g. Friedman and Schwartz (1963)), an endogenous reduction of money multiplier
(given a xed monetary base) leads to deation in our setting. However, in our setting
outside money is only an imperfect substitute for inside money. Intermediaries, either by
channeling funds through or by underwriting and thereby enabling rms to approach capital
markets directly, enable a better capital allocation and more economic growth. Hence, in our
setting monetary intervention should aim to recapitalize undercapitalized borrowers rather
than simply increase the money supply across the board. Another dierence is that our
approach focuses more on the role of money as a store of value instead of the transaction
role of money. Overall, our approach is closer in spirit to banking channel literature, see
e.g. Patinkin (1965), Tobin (1970), Gurley and Shaw (1955), Bernanke (1983) Bernanke and
1
If the interest rate is 0, then of course perpetual bonds would have an innite nominal price.
3
Blinder (1988) and Bernanke, Gertler and Gilchrist (1999).
2
Another distinct feature of our
setting is that our eects arise despite the fact that prices are fully exible. This is in sharp
contrast to the New Keynesian framework, in which a nominal interest rate cut also lowers
real rates and thereby induces households to consume more. Recently, Cordia and Wood-
ford (2010) introduced nancial frictions in the new Keynesian framework. In contrast, our
framework focuses on the redistributional role of monetary policy, a feature we share with
Scheinkman and Weiss (1986). In Kiyotaki and Moore (2008) money is desirable as it does
not suer from a resellability constraint, unlike capital in their model.
Within a three-period framework, Diamond and Rajan (2006) and Stein (2010) also
address the role of monetary policy as a tool to achieve nancial stability. More generally,
there is a growing macro literature which also investigated how macro shocks that aect
the balance sheets of intermediaries become amplied and aect the amount of lending and
the real economy. These papers include Bernanke and Gertler (1989), Kiyotaki and Moore
(1997) and Bernanke, Gertler and Gilchrist (1999), who study nancial frictions using a
log-linearized model near steady state. In these models shocks to intermediary net worths
aect the eciency of capital allocation and asset prices. However, log-linearized solutions
preclude volatility eects and lead to stable system dynamics. Brunnermeier and Sannikov
(2010) also study full equilibrium dynamics, focusing on the dierences in system behavior
near the steady state, and away from it. They nd that the system is stable to small shocks
near the steady state, but large shocks make the system unstable and generate systemic
endogenous risk. Thus, system dynamics are highly nonlinear. Large shocks have much
more serious eects on the real economy than small shocks. He and Krishnamurthy (2010)
also study the full equilibrium dynamics and focus in particular on credit spreads. For a
more detailed review of the literature we refer to Brunnermeier et al. (2010).
This paper is organized as follows. Section 2 informally describes the logical framework
around which we construct our model. Section 3 frames the ideas from Section 2 into a basic
model. Section 4 characterizes the equilibrium. Section 5 presents a computed example and
discusses equilibrium properties, including capital and money value dynamics, the amount
of lending through intermediaries, and the money multiplier. Section 6 introduces long-
term bonds and studies the eect of interest-rate policies as well as open-market operations.
Section 7 concludes.
2
The literature on credit channels distinguishes between the bank lending channel and the balance
sheet channel (nancial accelerator), depending on whether banks or corporates/households are capital
constrained. Strictly speaking our setting refers to the former, but we a agnostic about it and prefer the
broader credit channel interpreation.
4
2 Informal Description of the Economy
In this section we describe informally how we think about money and intermediation. The
goal is to explain the logic behind the main results and to lay out a general framework that
we implement in the next section.
The economy is populated by intermediaries and heterogeneous households. The distribu-
tion of productivity among households does not match the distribution of wealth. As a result,
productive households, whom we call entrepreneurs, need nancing to be able to manage
capital. In the absence of intermediaries there are extreme nancial frictions: unproductive
households with excess wealth cannot lend directly to nancially constrained entrepreneurs.
In the absence of money, these frictions lead to an extremely inecient allocation of capital,
in which each agent holds the amount of capital that is proportionate to his net worth. For
example, if the wealth of entrepreneurs is only 1% of aggregate wealth, then they can hold
only 1% of capital.
If household types are switching, then there can be a more ecient equilibrium with at
money. Assume that there is a xed supply of innitely divisible money. Even though it is
intrinsically worthless, in equilibrium money can have value by a mechanism which can be
related to the models of Samuelson (1958) and Bewley (1980).
3
Crucially, in order for money
to have value, enough agents should create demand for new savings through money to oset
the supply of money by agents who want to spend it to consume. In our model, this demand
stems from agents who suddenly become unproductive, and who want to exchange their
capital to hold money. If so, then these agents supply capital and demand money, agents
who stay unproductive supply money and demand output, and agents who stay productive
supply output and demand more capital. In equilibrium, the relative prices of capital, money
and output are determined so that all markets clear.
Money in equilibrium can lead to a more ecient allocation of capital. Even with ex-
treme nancial frictions that preclude borrowing and lending, the allocation of capital across
agents does not have to be proportional to their net worths. In fact, it may be possible for
entrepreneurs to hold all capital in the economy, while unproductive households hold money.
Nevertheless, when there is investment, the equilibrium with money is less ecient than
3
In Samuelson (1958), young agents are willing to save their wealth in money because they expect that
when they get old, they can trade money for consumption goods with the next generation of agents. In
Bewley (1980), agents are willing to accumulate money in periods when they have high endowment because
they expect to be able to trade money for consumption goods when their endowment is low, with agents
who have high endowment in that period.
5
the ecient outcome that arises in the absence of nancial frictions. Without borrowing and
lending, the entrepreneurs demand for capital is limited by their net worths. As a result,
even with money in equilibrium, capital becomes undervalued, leading to underinvestment
in capital. Furthermore, given low capital valuations, unproductive households may nd it
attractive to hold some capital, leading to an inecient allocation.
We introduce intermediaries who can mitigate the nancial frictions by facilitating lend-
ing from unproductive households to productive ones. Intermediaries can take deposits from
unproductive households to extend loans to entrepreneurs.
4
Importantly, when they fa-
cilitate the ow of funds from unproductive agents to entrepreneurs, intermediaries must
invariably be exposed to the risks of the projects they nance. They msut have some skin
in the game. The intermediaries ability to perform their functions depends on their risk-
bearing capacity. Because intermediaries are subject to the solvency constraint, their ability
to absorb risks depends on their net worths, and so after losses they are less able to perform
their functions.
Risk taking by intermediaries leads economic uctuations between a two polar cases: One
polar case looks like the benchmark without nancial frictions and in the second regime there
is no lending and hence money plays a much more important role. In the former regime,
banks create a large quantity of inside money by lending freely. Unproductive agents have
alternative ways to save other than holding outside money - they can hold deposits with
intermediaries (or entrepreneur equity). As a result, outside money has low value. At the
same time, easy nancing leads to a high price of capital and high investment.
If an aggregate macro shock causes intermediaries to suer losses, lending contracts,
causing entrepreneurs to reduce their demand for capital. As a result, the price of capital and
investment fall. At the same time, as the creation of inside money decreases, unproductive
households bid up the value of outside money to satisfy their demand for savings. This leads
to a collapse of the money multiplier and deation.
When deposits with intermediaries are denominated in money rather than output, then
deation increases the value of liabilities of intermediaries. Thus, intermediaries are doubly
hit: on the asset side because the value of capital that they nance decreases, and on the
liabilities side because the real value of their obligations goes up in value.
We construct a model that captures these dynamic eects in the next section. In Sec-
tion 4 we introduce in addition some long-term bonds and study how monetary policy and
4
They can also help entrepreneurs issue outside equity directly to households. Although we do not allow
for this possibility in our baseline model, we have explored it in an extension.
6
macroprudential policy can mitigate these adverse eects.
3 The Formal Baseline Model
We consider an innite-horizon economy populated by heterogeneous households and inter-
mediaries. Household types are denoted by , where could be an interval, or a nite
set. Higher types are more productive.
In our baseline model, which we present in this section, we assume that there is a xed
amount of gold in the economy, which serves as money. Gold is intrinsically worthless, and
the total quantity of gold is xed.
Household Production Technologies. The technology of household type generates
output at rate a

t
per unit of capital, where a

is the productivity parameter and

t
is
the rate of investment. Capital is measured in eciency units, and the quantity of capital
evolves according to
dk
t
k
t
= ((
t
)

) dt + d

t
. (3.1)
Function reects a decreasing-returns-to-scale investment technology, with (0) = 0,

>
0 and

< 0. That is, in the absence of investment, capital managed by household simply
depreciates at rate

. The term d

t
reects Brownian fundamental shocks to technology .
The technological shocks of types and

have covariance (,

), so that (, ) is
the volatility of

t
. We assume that a

weakly increases in type , while

decreases in .
Intermediation. Intermediaries can lend to productive households or invest in their
equity. In our model equity investment in household works as if the intermediary were
directly holding capital employed under the production technology of household , except for
an additional monitoring cost. We express the monitoring cost of equity nancing through
an increased depreciation rate by .
Thus, intermediaries can improve eciency in the economy in two ways: by channeling
funds to the most productive technologies, and by diversifying household risks on their
balance sheets. Intermediaries nance themselves borrowing from households.
Markets for Capital, Money and Consumption Goods. All markets are fully
liquid. All agents are small, and can buy or sell unlimited quantities of capital, money and
output in the market at any moment of time at current prices without making any price
impact. The aggregate quantity of capital in the economy is denoted by K
t
, and q
t
is the
7
price of one unit of capital in the units of output. The total quantity of gold is normalized
to one, and the value of all gold in the units of output is denoted by P
t
.
The aggregate amount of capital K
t
depends on aggregate investment, the allocation of
capital among the dierent technologies, and technology shocks. The price of capital q
t
and
the value of money P
t
are determined endogenously from supply and demand.
Net Worths and Balance Sheets. The total net worth of all agents is
q
t
K
t
+ P
t
.
The main focus of our model is on the net worth of intermediaries is N
t
< q
t
K
t
+ P
t
. The
net worth of the intermediary sector N
t
relative to the size of the economy K
t
determines
the risk-taking capacity of intermediaries, the amount of nancing available to productive
technologies and the creation of inside money.
To keep the model tractable, we ignore any eects that arise through changing wealth
distribution among households. Specically, we assume that wealth q
t
K
t
+P
t
N
t
is always
distributed among households with an exogenous density (), with
_

() d = 1.
That is, even though shocks and heterogeneous productivity have dierent eects on the
net worth of dierent households, we assume that household types are completely transient
- they switch fast enough to eliminate any temporary wealth accumulation eects. At the
beginning of each period, each household gets randomly reassigned to a new type according
to the probability distribution ().
Each household chooses how to allocate wealth between its own productive technology
and money. Households may borrow and become levered, allocating a negative portfolio
weight to money. Intermediaries invest in technologies of a range of household types . We
denote the equilibrium allocation of capital across technologies, and between households and
intermediaries through functions
t
() and
t
() such that
_

t
() d +
_

t
() d = 1.
Function
t
() describes the density of the allocation of capital across households, and
t
(),
the technology portfolio of intermediaries. Given the allocation of capital, the aggregate
8
amount of capital in the economy follows
dK
t
K
t
=
__

(
t
() +
t
())g

(q
t
) d
_

t
() d
_
. .

K
t
dt +
_

t
(
t
() +
t
()) d
. .
d
K
t
. (3.2)
Optimal Investment. Given the current price of capital q
t
, the optimal investment
rate
t
is identical across all agents. It solves
max

()q
t
,
where () is rate at which new capital is created and q
t
is the price of capital. Since

< 0,
the solution is uniquely pinned down by the rst-order condition

()q
t
= 1. We denote the
solution by (q
t
), and the resulting net output rates and capital growth rates by
c

(q
t
) = a

(q
t
) and g

(q
t
) = ((q
t
))

.
Returns on Capital and Money. We denote the law of motion of the price of capital
q
t
by
dq
t
q
t
=
q
t
dt + d
q
t
. (3.3)
As a result, household of type earns return on capital of
dr

t
=
c

(q
t
)
q
t
. .
dividend yield
dt + (g

(q
t
) +
q
t
+ Cov(d

t
, d
q
t
))
. .
expected capital gains rate
dt + d

t
+ d
q
t
. .
risk
.
The capital gains rate and risk, which equal d(k
t
q
t
)/(k
t
q
t
), and are derived from (3.1) and
(3.3) using Itos lemma.
Intermediaries get a return of
dr

t
dt (3.4)
from investing in technology . The return on money is given by the law of motion of P
t
,
dr
M
t
=
dP
t
P
t
=
M
t
dt + d
M
t
. (3.5)
Preferences and Utility Maximization. All agents have logarithmic utility with a
9
common discount rate of . Logarithmic utility has two convenient properties that signi-
cantly simplify the analysis of our model.
1. Any agent with logarithmic utility consumes times his net worth.
2. The solution of the optimal portfolio choice problem of an agent with log utility has
a very simple characterization. On the margin, any small change in asset allocation
around the optimum has
marginal expected return = Cov(marginal risk, net worth risk at optimum). (3.6)
In this equation, net worth risk is measured per dollar of net worth. We demonstrate
further the mechanics of applying this condition.
Importantly, the consumption and investment decisions of an agent with logarithmic util-
ity are completely myopic, irrespective of future investment opportunities. The solution
procedure for more general preferences specications is more complicated, and requires the
introduction of the agents value functions and the Bellman equation.
In addition, to control the size of the intermediary sector in our model, we assume that
intermediaries may retire. Intermediaries have a dicult job, which requires eort. Upon
retirement, they relax and receive an immediate utility boost of x. We denote the retirement
rate of experts by N
t
d
t
.
The Evolution of N
t
. Since we know the portfolio of technologies
t
(), in which inter-
mediaries invest, the return of each technology (3.4) and the promised return to depositors
(3.5), the law of aggregate intermediary net worth is
dN
t
= N
t
dr
M
t
C
t
dt + q
t
K
t
_

t
()(dr

t
dt dr
M
t
) d N
t
d
t
. (3.7)
In this equation, C
t
is the aggregate rate of intermediary consumption.
Equilibrium Denition. An equilibrium is characterized by price processes q
t
and P
t
,
the allocation
t
() and
t
() of aggregate capital K
t
, consumption rates of all agents and the
rate of intermediary retirement, which map any history of Brownian shocks {Z
s
, s [0, t]}
into the state of the economy at time t. The following conditions have to hold
(i) All markets, for capital, money and consumption goods, clear
10
(ii) Intermediaries choose consumption, investment () and retirement to maximize utility
and
(iii) Households of each type choose consumption and the allocation of wealth between
money and capital to maximize utility
Scale Invariance and Markov Equilibria. Intuitively, because our setting is scale-
invariant, the severity of nancial frictions in our economy is quantied by the ratio of expert
net worth to the size of the economy,

t
=
N
t
K
t
.
We expect two economies in which the ratio
t
is the same look like scaled versions of
one another, and look for a Markov equilibrium with the state variable
t
. In a Markov
equilibrium, the price of capital q
t
= q(
t
) and the allocation of capital {
t
(),
t
()} =
{(
t
, ), (
t
, )} are functions of
t
, while the value of money is proportional to K
t
, i.e.
P
t
= p(
t
)K
t
.
The laws of motion of P
t
and p
t
= p(
t
) are related by
dr
M
t
=
dP
t
P
t
=
_

p
t
+
K
t
+ Cov(d
p
t
, d
K
t
)
_
. .

M
t
dt + d
p
t
+ d
K
t
. .
d
M
t
, where
dp
t
p
t
=
p
t
dt + d
p
t
. (3.8)
4 Equilibrium Conditions
In this section we derive equilibrium conditions, and develop a method to compute equilibria
numerically. We focus on the following conditions from our denition of equilibria:
(i) The market-clearing conditions - two equations guarantee that all three markets clear,
by Walras law
(ii) Intermediaries borrow money to gain exposure to each technology optimally
(iii) Each household type chooses the allocation of wealth between money and its own
technology optimally
Besides these conditions, we embed the optimal consumption rates into the market-clearing
conditions directly, and embed the optimal retirement rate of intermediaries into the law of
motion of the state variable
t
.
11
Market-Clearing Conditions. The condition that the market for consumption goods
has to clear is
(q(
t
) + p(
t
)) =
_

((
t
, ) + (
t
, )) a

d (q(
t
)), (4.1)
since (q
t
+ p
t
)K
t
is aggregate net worth of all agents, and the right hand side of (4.1) is net
equilibrium output per unit of capital.
The market-clearing condition for capital is just
_

(
t
, ) d +
_

(
t
, ) d = 1. (4.2)
Intermediaries Optimal Portfolio Choice. The optimal portfolio choice conditions
are based on equation (3.6). The marginal expected return rate from an investment in
technology , nanced by borrowing money, is E[dr

t
dt dr
M
t
]. The marginal risk of
this investment is d

t
+ d
q
t
d
M
t
, and, from (3.7), the risk of intermediary net worths is
d
N
t
= d
M
t
+
q
t

t
_

t
(

)(d

t
+ d
q
t
d
M
t
) d

. (4.3)
Therefore, intermediaries choose their portfolios optimally if and only if
E[dr

t
dt dr
M
t
] Cov(d

t
+ d
q
t
d
M
t
, d
N
t
), (4.4)
with equality if intermediaries invest a nonzero amount in technology , i.e. (
t
, ) > 0.
Households Optimal Portfolio Choice. Household of type chooses how to allocate
its wealth between its own technology and money. The marginal expected return from an
additional allocation to capital is E[dr

t
dr
M
t
], while the marginal risk of this investment is
d

t
+d
q
t
d
M
t
. To determine the risk of this households net worth, note that if holds capital
with value (
t
, )q
t
K
t
on the net worth of ()(q
t
+ p
t

t
)K
t
. Therefore, the household
faces risk per unit of net worth of
d
M
t
+
(
t
, )q
t
()(q
t
+ p
t

t
)
(d

t
+ d
q
t
d
M
t
).
12
The optimal portfolio choice condition of household is
E[dr

t
dr
M
t
] Cov
_
d

t
+ d
q
t
d
M
t
, d
M
t
+
(
t
, )q
t
()(q
t
+ p
t

t
)
(d

t
+ d
q
t
d
M
t
)
_
, (4.5)
with equality if the household invests a positive fraction of wealth in its technology, i.e.
(
t
, ) > 0.
The Law of Motion of
t
. The drift and volatility of q
t
and p
t
that enter equilibrium
conditions (4.4) and (4.5) depend on the law of motion of the state variable
t
. Proposition
1 allows us to reduce the equilibrium conditions to a system of dierential equations for q()
and p(), together with the portfolio choice processes {(
t
, ), (
t
, )}.
Proposition 1. In equilibrium
t
follows
d
t

t
= (
p
t
+ Cov(d

t
d
p
t
, d

t
)) dt + d

t
d
t
, where d

t
= d
N
t
d
K
t
. (4.6)
Proof. From (4.4) and (4.3),
q
t

t
_

t
()(dr

t
dtdr
M
t
)d =
q
t

t
_

t
()Cov(d

t
+d
q
t
d
M
t
, d
N
t
)d = Cov(d
N
t
d
M
t
, d
N
t
).
Therefore, from (3.7) and (3.8), the law of motion of N
t
is given by
dN
t
N
t
=
_

p
t
+
K
t
+ Cov(d
p
t
, d
K
t
) + Cov(d
N
t
d
M
t
, d
N
t
)
_
dt + d
N
t
d
t
. (4.7)
Using (3.2) and Itos lemma,
d(1/K
t
)
1/K
t
= (Var(d
K
t
)
K
t
) dt d
K
t
.
Since
t
= N
t
(1/K
t
), we have
d
t

t
=
dN
t
N
t
+
d(1/K
t
)
1/K
t
Cov(d
N
t
, d
K
t
) dt =
_

p
t
+ Cov(d
p
t
+ d
K
t
d
N
t
, d
K
t
) + Cov(d
N
t
d
M
t
, d
N
t
)
_
dt + d
N
t
d
K
t
d
t
,
which simplies to (4.6).
13
Because the decision to retire is a real-option problem for experts, they exercise the
option when
t
reaches a suciently high critical level

, at which competition among

experts makes them indierent between retiring or not. As a result, the state variable
t
follows (4.6) with d
t
= 0 when
t
<

. The process
t
makes
t
reect at

.
Dierential Equations and Boundary Conditions. Equilibrium conditions (4.1),
(4.2), (4.4) and (4.5) can be converted into a system of dierential equations for q(), p(),
as well as {(
t
, ), (
t
, )}, using Itos lemma. Terms d
q
t
, d
p
t
, as well as
q
t
and
p
t
, can
be expressed in terms of the drift and volatility of
t
as follows
d
q
t
=
q

(
t
)
t
q(
t
)
d

t
, d
p
t
=
p

(
t
)
t
p(
t
)
d

t
,
_
1
p

(
t
)
t
p(
t
)
_

p
t
=
p

(
t
)
t
p(
t
)
__
1
p

(
t
)
t
p(
t
)
_
(

)
2

_
+
1
2
p

(
t
)
t
p(
t
)
(

)
2
,

q
t
=
q

(
t
)
t
q(
t
)
_

p
t
+
_
1
p

(
t
)
t
p(
t
)
_
(

)
2
_
+
1
2
q

(
t
)
t
q(
t
)
(

)
2
,
where (

)
2
= Var(d

t
).
The boundary conditions are as follows. The value of

is uniquely pinned down by

the parameter x, the utility boost that intermediaries receive upon retirement. In order to
prevent abnormal prots or losses at the reection point

, we need that
d
q
t
= d
p
t

p

)
p(

)
=
q

)
q(

)
.
The level of q(

) has to be chosen to match the autarky boundary condition at = 0, which

corresponds to the equilibrium without intermediaries.
5 An Example
In this version of the paper, we present an example from a simplied version of the model
with three types = {
L
,
M
,
H
}. The distribution of wealth among households is given by
(
L
) = 65%, (
M
) = 35% and (
H
) = 0. All three production technologies lead the same
gross output rate a

= 1, but dierent depreciation rates that satisfy

L
= ,

M
= 3%
and

H
+ = 0%. Note that, since the most productive households have zero net worth,
we only need to know their productivity net of the monitoring cost. The monitoring cost
14
is large enough so that intermediaries do not want to invest in type
M
technology. The
investment function is (i) =

0.02i, and shocks to the agents technologies are given by

d
M
t
= 0, d
H
t
= dZ
t
+

dZ

t
,
where the Brownian motion Z
t
represents the aggregate shocks, and Z

t
represent idiosyn-
cratic shocks, which are independent and fully diversiable across individual agents.
The three household types play very specic distinct roles in this example. The low
types create a demand for money. The middle types, who do not get nancing from the
intermediaries, exist to ensure that the average of the shocks that hit the economy is distinct
from the average of the shocks that hit intermediary assets. Intermediaries invest in projects
with higher and more concentrated risk compared the risk of their liabilities, i.e. money,
which in the long run has the same risk as the economy as a whole, that is, d
K
t
.
5
Regarding preferences, we set the discount rate to = 5%, and select the intermediaries
eort cost x to induce them to retire at

= 2.1.
Figure 1 presents market prices of money and capital, as well as capital allocation, inter-
mediary leverage and the volatility of
t
.
As we expected, the price of capital increases with while the price of money decreases
with . Intermediaries are able to hold more capital for higher , i.e. (,
H
) is increasing
in . At the same time, when is low, then lower price of capital encourages intermediaries
to take on greater leverage. As 0, the prices of capital and money converge to q(0) =
(
M
)a

M
/ = 7 and p(0) = (
L
)a

M
/ = 13, since
t
stays at zero if it ever reaches 0.
Both q

() and p

() are large near = 0, which creates signicant amplication in the

depressed region. A negative macro shock causes the price of capital to fall and the price of
money to rise, hitting intermediaries on both asset and liability sides of their balance sheets.
As a result, the volatility of
t
, which reects endogenous risk, is high near = 0.
Figure 2 displays the expected rate of economic growth, investment, the volatility of
capital and money, and
q
t

p
t
, which enters the expected return of a levered position in
capital. As expected, economic growth and investment are higher when
t
is larger. The
bottom left panel shows the reaction of the values of capital and money to aggregate shocks.
Capital becomes cheaper after a negative macro shock as expert demand for capital decreases
following a contraction of their balance sheets. On the other hand, a negative macro shock
5
If there were only two household types, unproductive and productive, then there would be an equilibrium
in which q
t
and p
t
are constant, and both intermediary assets and liabilities are perfectly hedged with risk
d
K
t
.
15
Figure 1: Asset prices, capital allocation, leverage and

t
in dynamic equilibrium.
Figure 2: Economic growth, investment, asset price volatilities and returns in equilibrium.
16
generally increases the value of money, except for
t
near

. As intermediary balance sheets

contract, the amount of inside money they create decreases and, to oset this eect, the
value of outside money has to rise. Note also that the volatility of a levered position in
capital is highest when is low. This volatility is due to endogenous risk, which is created as
the macro shocks get amplied through the intermediaries balance sheets. High volatility
leads to a high required risk premium. Hence,
q
t

p
t
is highest when volatility due to
endogenous risk is highest.
Figure 3: Fundamental components of the returns on capital and money.
The impact of
q
t
and
p
t
on the returns of capital and money can be signicant. For

t
between 0 and 0.5, these valuation eects easily overwhelm the eects of the changes in
fundamentals on the return on money and capital. Recall that the fundamental return on
capital stems from output and investment while the fundamental return on money stems from
economic growth, i.e. money buys a xed fraction of the economy in the long run. Figure
3 displays the fundamental components of the return on capital held by intermediaries and
type
M
households, as well as money. This component of the return on capital decreases
in as capital becomes more expensive. At the same time, the return on money increases
with the expected rate of economic growth.
Importantly, Figure 4 displays various money quantities in our equilibrium. The top
panel of Figure 4 exhibits the value of all outside money p(), as well as the total value of
money and the value of inside money. As increases, intermediaries invest more and create
more inside money. As the value of outside money decreases at the same time, the money
multiplier expands, as shown on the bottom panel of Figure 4.
17
Figure 4: Inside and outside money, and the money multiplier.
6 Long-Term Bonds and Monetary Policy
In this section we analyze how a central bank can aect the equilibrium through monetary
policy that involves setting short-term interest rates and using open-market operations to
change the composition of monetary instruments. As a theoretical exercise, we would like
to isolate monetary policies from more general scal policies. We do not consider policies
that involve government spending, taxation, or explicit redistribution. Rather, we consider
a central bank that only has the authority to print money and can issue perpetual long-
term bonds, securities that promise a perpetual stream of monetary payments in the future.
Interest on long-term bonds can be nanced, again, by printing money or selling more long-
term bonds.
Unlike in our benchmark model, with active monetary policy the amount of money in
the economy is no longer xed. The quantity of money changes over time, because anybody
(households or intermediaries) can deposit money with the central bank to earn a nominal
interest rate of r
t
. In addition, the central bank issues perpetual long-term bonds and sells
them in the open market in exchange for money. The central bank can also print money
to repurchase some of the bonds. We summarize the policy that aects the composition of
18
outstanding monetary instrument through a process b
t
, where b
t
K
t
is the real market value
of all long-term bonds outstanding. Thus, two policy instruments - short-term interest rates
and open-market operations - are summarized by a pair of processes {r
t
, b
t
}.
For a given policy, we denote the equilibrium nominal price of long-term bonds per unit
coupon rate by B
t
. For example, if r
t
= r for all t, then the nominal value of a bond that
pays 1 in perpetuity is B
t
= 1/r. If instead in downturns the central bank lowers r
t
, then we
expect B
t
to increase. We denote the equilibrium law of motion of B
t
by
dB
t
B
t
=
B
t
dt +
B
t
. (6.1)
With long-term bonds, the total real wealth in the economy is given by
(q
t
+ p
t
+ b
t
)K
t
,
where p
t
K
t
still denotes the real market value of money and b
t
K
t
is the real market value of
long-term bonds.
The stochastic laws of motion of both p
t
and b
t
,
db
t
b
t
=
b
t
dt +
b
t
, (6.2)
reect both the changes in the real value of money and long-term bonds, as well as the
open-market operations conducted by the central bank. One has to take those central bank
transactions into account to gure out exactly how the total values of money and bonds
outstanding are related to the returns that money and bonds earn. The following proposition
disentangles the algebra of open-market operations.
Proposition 2. Given our notation, the real return on money can be expressed as
dr
M
t
= (
p
t
+
b
t
+
K
t
+Cov(d
p
t
+d
b
t
, d
K
t
)) dt +d
p
t
+d
b
t
+d
K
t

b
t
p
t
+ b
t
(dr
B
t
dr
M
t
), (6.3)
where the dierence between the real returns on bonds and money is
dr
B
t
dr
M
t
=
_
1/B
t
r
t
+
B
t
+ Cov(d
B
t
, d
M
t
)
_
dt + d
B
t
. (6.4)
Proof. To see why equation (6.4) holds, consider an agent who borrows money to buy bonds.
This agent receives interest on bonds of 1/B
t
per dollar investment, and has to pay interest
19
r
t
on borrowed money. The agent is also exposed to the uctuations of the price of bonds
relative to money,
B
t
dt + d
B
t
, but is perfectly hedged to the uctuations of the real value
of money. The uctuations in the value of money add to the agents return only insofar as
they are correlated to the uctuations of the nominal price of bonds, leading to the term
Cov(d
B
t
, d
M
t
).
Regarding (6.3), note that an investment into all monetary instruments in the economy
(money and long-term bonds) earns a return of
d((p
t
+ b
t
)K
t
)
(p
t
+ b
t
)K
t
= (
p
t
+
b
t
+
K
t
+ Cov(d
p
t
+ d
b
t
, d
K
t
)) dt + d
p
t
+ d
b
t
+ d
K
t
. (6.5)
Of this portfolio, fraction b
t
/(p
t
+ b
t
) is invested in bonds, so
d((p
t
+ b
t
)K
t
)
(p
t
+ b
t
)K
t
= dr
M
t
+
b
t
p
t
+ b
t
(dr
B
t
dr
M
t
). (6.6)
Combining (6.5) and (6.6), we get (6.3).
Equilibrium Equations. We are now ready to write down equilibrium equations, which
are analogous to those in Section 4. The only dierence is that now, agents are free to hold
not only money, but also bonds. To keep things simple, we consider monetary policies that
are Markov in the state variable
t
. We are particularly interested in policies that lower the
short-term interest rate in downturns and raise it in booms.
We denote the fraction of bonds allocated to intermediaries by
B
t
, and the density of bond
holdings across household types by
B
t
(). Then the market-clearing condition for bonds is

B
t
+
_

B
t
() d = 1. (6.7)
While the market-clearing condition for capital is still (4.2), the market-clearing condition
for output changes slightly to
(q
t
+ p
t
+ b
t
) =
_

(
t
() +
t
()) a

d (q
t
). (6.8)
To write down the optimal portfolio choice conditions, note that intermediaries are exposed
20
to the risk of
d
N
t
= d
M
t
+
q
t

t
_

t
(

)(d

t
+ d
q
t
d
M
t
) d

+
b
t

B
t
d
B
t
(6.9)
per unit of net worth, while household of type face the risk
d
N,
t
= d
M
t
+

t
()q
t
()(q
t
+ p
t

t
)
(d

t
+ d
q
t
d
M
t
) +

B
t
()b
t
()(q
t
+ p
t

t
)
d
B
t
. (6.10)
The optimal portfolio choice equations are
E[dr

t
dt dr
M
t
] Cov(d

t
+ d
q
t
d
M
t
, d
N
t
), (= if
t
() > 0), (6.11)
E[dr
B
t
dr
M
t
] Cov(d
B
t
, d
N
t
), (= if
B
t
> 0), (6.12)
E[dr

t
dr
M
t
] Cov
_
d

t
+ d
q
t
d
M
t
, d
N,
t
_
, (= if
t
() > 0), (6.13)
and E[dr
B
t
dr
M
t
] Cov
_
d
B
t
, d
N,
t
_
, (= if
B
t
() > 0). (6.14)
The Eects of Monetary Policy. In our model, monetary policy can mitigate down-
turns in two ways. First, by increasing the money supply in downturns, reducing the dea-
tionary spiral. Second, by redistributing wealth towards intermediaries, whose holdings of
long-term bonds increase in value due to cuts in interest rates. Both of these eects appear
when the central bank lowers r
t
when
t
goes down.
To see why the deationary spiral is mitigated by interest rate cuts in downturns, note
that the demand for money stems from the unproductive households desire to save. In
the baseline model, when the intermediary sector creates less inside money in downturns,
household demand for outside money goes up. That raises the value of outside money,
leading to deation. With a central bank, unproductive households can use both money
and long-term bonds to save. The central bank can counteract the contraction of the money
multiplier in downturns by cutting interest rates, and thereby increasing the value of long-
term bonds. That satises the households demand for savings, and eases the deationary
pressure.
The redistributing eects arise because intermediaries can hold long-term bonds to hedge
losses in downturns. The increase in the value of long-term bonds osets some of the losses
that intermediaries suer on their assets. Thus, interest rate cuts in downturns recapitalize
intermediaries. Also, due to o-setting risks of capital and long-term bonds, intermedi-
21
aries can absorb more capital risk in downturns, so they can lend more and create more
inside money. Of course, policies that help intermediaries in downturns may create greater
incentives for risk-taking ex-ante.
7 Conclusion
// to be completed //
8 Bibliography
Bernanke, B. (1983) Non-Monetary Eects of the Financial Crisis in the Propagation
of the Great Depression, American Economic Review, Vol. 73, 257-276.
Bernanke, B. and A. Blinder (1989) Credit, Money, and Aggregate Demand, American
Economic Review, Papers and Proceedings, Vol. 78, No. 2, 435-439.
Bernanke, B. and M. Gertler (1989) Agency Costs, Net Worth, and Business Fluctua-
tions, American Economic Review, Vol. 79, No. 1, 14-31.
Bernanke, B., M. Gertler and S. Gilchrist (1999) The Financial Accelerator in a Quan-
titative Business Cycle Framework, Handbook of Macroeconomics (Chapter 21), 1341-1393.
Bewley, T. (1980) The Optimum Quantity of Money, in Models of Monetary Eco-
nomics, eds. J. Kareken and N. Wallace. Minneapolis, Minnesota: Federal Reserve Bank.
Brunnermeier, M. K., T. Eisenbach, and Y. Sannikov (2010) Macroeconomics with
Financial Frictions: A Survey, in Proceedings of the Econometric Society World Congress
2010 in Shanghai, eds. E. Dekkel, Cambridge University Press.
Brunnermeier, M. K. and Y. Sannikov (2010) A Macroeconomic Model with a Financial
Sector, working paper, Princeton University.
Cagan, P. (1979) Persistent Ination: Historical and Policy Essays, Columbia university
Press.
Cordia, V. and M. Woodford (2010) Credit Spreads and Monetary Policy, working
paper, Columbia University.
Diamond, D. W. (1984) Financial Intermediation and Delegated Monitoring, The Re-
view of Economic Studies, 51(3), 393414.
Diamond, D. and R. Rajan (2006) Money in a Theory of Banking, American Economic
Review, March 2006.
22
Friedman, M. and A. J. Schwartz. (1963) A Monetary History of the United States,
1867-1960, Princeton University Press.
Gurley, J and E. Shaw (1955) Financial Aspects of Economic Development, American
Economic Review, September 2006, 515-538.
He, Z., and A. Krishnamurthy (2010) Intermediary Asset Prices, working paper, North-
western University.
He, Z., and A. Krishnamurthy (2010) A Model of Capital and Crises, forthcoming in
Review of Economic Studies.
Holmstrom, B., and J. Tirole (1997): Financial Intermediation, Loanable Funds, and
the Real Sector, The Quarterly Journal of Economics, 112(3), pp.663691.
Kiyotaki, N. and J. Moore (1997) Credit Cycles, Journal of Political Economy, Vol.
105, No. 2, 211-48.
Kiyotaki, N. and J. Moore (2008) Liquidity, Business Cycles, and Monetary Policy,
working paper, Princeton University.
Patinkin D. (1965) Money, Interest and Prices: An Integration of Monetary and Value
Theory, Harper & Row NY.
Samuelson, P. A. (1958) An exact consumption loan model of interest, with or without
the social contrivance of money Journal of Political Economy 66, 467-82.
Stein, J. (2010) Monetary Policy as Financial-Stability Regulation, working paper,
Harvard University.
Scheinkman, J. A. and L. Weiss (1986) Borrowing Constraints and Aggregate Economic
Activity, Econometrica, Vol. 54, 23-45
Tobin, J. (1969) A General Equilibrium Approach to Monetary Theory, Journal of
Money, Credit and Banking, Vol. 1, No. 1, 15-29.
23